The Spine Journal 4 (2004) 373–378

Clinical Studies

Injury severity as primary predictor of outcome in acute spinal cordinjury: retrospective results from a large multicenter clinical trial

William P. Coleman, PhDa, Fred H. Geisler, MD, PhDb,*aWPCMath, 703 West Ferry Street, C-20, Buffalo, NY 14222, USA

bChicago Institute of Neurosurgery and Neuroresearch and RUSH University, 4501 N. Winchester, 3rd Floor, Chicago, IL 60640, USA

Received 4 June 2003; accepted 29 December 2003

Abstract BACKGROUND CONTEXT: The prognostic value of injury severity and of anatomical regionin acute spinal cord injury is strong, making it hard to evaluate other indicators or assess improvementwithout considering them.PURPOSE: This study documents issues and suggests a practical way to stratify.STUDY DESIGN/SETTING: Retrospective analysis of data prospectively collected for the multi-center trial of GM-1.PATIENT SAMPLE: A total of 760 patients were recruited at 28 centers in North America. Injurieswere rostral to T10 and left at least one leg with an American Spinal Injury Association (ASIA)motor score less than 15 of 25. Patients were assessed at baseline using the ASIA Impairment Scale(AIS): Grade A, Grade B, and Grades C and D (combined). They were divided by injury region:cervical or thoracic.OUTCOME MEASURES: The endpoint was marked recovery (MR), defined as improvement ofat least two grades from AIS at baseline to Modified Benzel Scale at Week 26. Other endpointswere changes in ASIA Motor, in light touch, and in pin prick scores.METHODS: Data were verified onsite by a central team of monitors, the database was checkedand standard statistical techniques were applied.RESULTS: Recruitment was uneven. In 760 patients, 579 injuries were cervical, and 482 werecomplete. There were few incomplete thoracic injuries. The cervical group had more MR thanthe thoracic group (37.2% vs 15.9%, p�.0001). AIS Groups C and D had (p�.0001) more MR(84.0%) than Group B (46.6%), which recovered more than Group A (12.8%). The cervical grouphad an advantage in MR because it had more patients with AIS B, and still more AIS C and D.Within AIS Group A, the cervical subgroup had (p�.02) higher MR (15.5%) than the thoracic one(7.0%), but MR was nearly equal in the B and CD groups. This suggested a new stratificationvariable, “injury region/severity,” to distinguish cervical (n�332, MR�15.5%) and thoracic (n�150,MR�7.0%) injuries within AIS A, but not in AIS B (n�131, MR�46.6%) or AIS CD (n�147,MR�84.1%). This variable is a significant predictor of MR (p�.0001).CONCLUSIONS: AIS severity was the strongest predictor. Anatomical region was also strong butconfounded with the severity effect, because the cervicals had fewer complete injuries, and becausethe cervical complete group did better than thoracic complete. The injury region/severity variable keepsthe strong prognostic value of using both region and severity, but is simpler and more statisticallyeconomical. � 2004 Elsevier Inc. All rights reserved.

Keywords: Acute spinal cord injury; Severity; Prognosis; Cervical; Thoracic; Outcome; ASIA Impairment Scale

FDA device/drug status: investigational/not approved (Sygen).Nothing of value received from a commercial entity related to this

research.* Corresponding author. Illinois Neuro-Spine Center, 2020 Ogden Ave.,

Suite 335, Aurora, IL 60504. Tel.: (630) 236-4303; fax: (630) 236-4303.E-mail address: fgeisler@concentric.net (F.H. Geisler)

1529-9430/04/$ – see front matter � 2004 Elsevier Inc. All rights reserved.doi:10.1016/j.spinee.2003.12.006

Introduction

The prognostic value of severity and of anatomical regionin acute spinal cord injury is strong, making it statisticallydifficult to evaluate treatment or other prognostic factors orto assess improvement without accounting for them.

It is possible, in principle, to compensate statisticallyeither by prospectively stratifying for these features or

W.P. Coleman and F.H. Geisler / The Spine Journal 4 (2004) 373–378374

Table 1Definition of American Spinal Injury Association Impairment Scale

Grade Description

A No motor or sensory function is preserved in the sacralsegments S4–S5

B Sensory but no motor function is preserved below theneurological level and extends through the sacral segmentsS4–S5

C Motor function is preserved below the neurological level,and the majority of key muscles below the level have amuscle grade less than 3.

D Motor function is preserved below the neurological level,and the majority of key muscles below the level have amuscle grade greater than or equal to 3.

E Motor and sensory function are normal.

by retrospectively modeling them. However, such methodsrequire skill and judgment. They introduce complexity andthe potential for muddying the results: the statistical accuracyof a study has to be divided among these factors in additionto those that are the main interest.

To make this problem more difficult, recruitment isuneven. Using the entry criteria of the multicenter GM-1trial, which was designed to study severe spinal cord injury,there are more cervical injuries than thoracic ones and morecomplete ones than incomplete. This presents a danger that,

Table 2Definition of the Modified Benzel Scale

EquivalentGrade Description AIS grade(s)

I No motor or sensory function is preserved Ain the sacral segments S4–S5

II Sensory but no motor function is preserved Bin the sacral segments S4–S5

III Motor function is preserved below the Cneurological level, and the majority of keymuscles below the neurological level havea muscle grade less than 3. Unable to walk.

IV Unable to walk. Some functional motor control (C*), Dbelow the level of injury that is significantlyuseful (assist in transfers, etc) but that is notsufficient for independent walking.

V Limited walking. Motor function allows walking Dwith assistance or unassisted, but significantproblems secondary to lack of endurance orfear of falling limit patient mobility. (Must beable to ambulate at least 25 feet.)

VI Unlimited walking. Ambulatory without Dassistance and without significant limitationsother than one or both of the following:difficulties with micturition, slightly dysco-ordinated gait. (Must be able to ambulate atleast 150 feet without a helper.)

VII Neurologically intact with the exception of mini- (D*), Emal deficits that cause no functional difficulties.(Must have a neurologically normal gait andbe able to walk without assistanceor assistive devices.)

AIS�American Spinal Injury Association Impairment Scale.

Table 3Definition of marked recovery

Modified Benzel Score at 26 Weeks

AIS grade at baseline I II III IV V VI VII

A MR MR MR MR MRB MR MR MR MRC MR MR MRD MR MR

AIS�American Spinal Injury Association Impairment Scale; MR�

marked recovery.

for example, authors or readers of studies heavily weightedin the cervical injuries might assume the results apply equallyto the less-represented thoracic ones.

This retrospective study documents the issues and sug-gests a practical way to stratify.

Methods

For the GM-1 multicenter clinical trial, a total of 760evaluable patients were recruited at 28 neurotrauma centersin North America during a 5-year period from April 1992 toJanuary 1997 [1–6]. In order to ensure they had an injury thatwas severe but still a pure spinal cord injury with potential forrecovery, the trial required the injury to be rostral to T10 andto leave at least one lowerextremity with an ASIA motor score[7] less than 15 of 25, and it excluded anatomic transections.Patients were required to have received the National AcuteSpinal Cord Injury Study 2 (NASCIS-2) dose of methylpred-nisolone sodium succinate (MPSS) for 24 hours, startingwithin 8 hours of injury.

Patients had an emergency room evaluation and then hada detailed baseline evaluation just before the first dose ofstudy medication, which began no later than 72 hours afterthe injury. The baseline examinations in this study wereperformed well after the initial trauma evaluation and resus-citation. This delayed examination is more reliable and com-plete than can be performed in the emergency room as partof the initial trauma evaluation and treatment.

Patients were divided by injury region: cervical or tho-racic. They were assessed at baseline using the AmericanSpinal Injury Association (ASIA) Impairment Scale (AIS),as defined in Table 1. For analysis, Grades C and D werecombined. Patients were assessed at Week 26 using the Mod-ified Benzel Scale, defined in Table 2, which expands theAIS into finer gradations.

Table 4Recruitment by spinal region and injury severity

AIS A AIS B AIS CD Total

Cervical 332 113 134 579Thoracic 150 18 13 181Total 482 131 147 760

The Fisher’s exact test significance level for this table is p�.0001.AIS�American Spinal Injury Association Impairment Scale.

W.P. Coleman and F.H. Geisler / The Spine Journal 4 (2004) 373–378 375

Table 5Potential predictors of marked recovery at Weeks 8 and 26

Week 8 Week 26

Category Group N MR p N MR p

AIS severity A 482 6.0% �.0001 460 12.8% �.0001B 131 21.4% 118 46.6%CD 147 56.5% 138 84.0%

Anatomical region Cervical 579 21.8% �.0001 546 37.2% �.0001Thoracic 181 7.7% 170 15.9%

Central cord Suspect 69 55.1% �.0001 66 83.3% �.0001Doubt or unknown 691 14.8% 650 26.9%

Type Stable 72 34.7% .0006 67 58.2% �.0001Fracture dislocation 688 16.7% 649 29.4%

Had spine surgery? Yes 600 18.5% 1.00 566 31.3% .38No 160 18.1% 150 35.3%

Surgical timing group �8 hours 25 24.0% .21 23 47.8% .088�x�24 94 23.4% 90 37.8%24�x�72 174 13.8% 161 24.8%72�x�240 195 21.0% 185 33.0%240�x 112 16.1% 107 29.0%

Tertiary care admit Direct 291 18.6% 1.00 276 33.0% .74Indirect 469 18.3% 440 31.6%

Age �30 years 367 17.4% .51 348 31.6% .81�30 393 19.3% 368 32.6%

MPSS timing �3 hours 469 18.6% 1.00 445 31.9% .87�3 hours 289 18.3% 269 32.7%

Treatment arm GM-1 430 21.6% .01 404 32.9% .63Placebo 330 14.2% 312 31.1%

AIS�American Spinal Injury Association Impairment Scale; MPSS�methylprednisolone sodium succinate; MR�marked recovery.

The primary endpoint was marked recovery (MR), whichis defined in Table 3 as an increase of at least two gradesfrom AIS at baseline to Modified Benzel Scale at Week 26.This reflects a degree of clinical improvement that is readilyapparent to patients, families and care providers.

Other endpoints were changes in ASIA motor, in lighttouch and in pin prick score. The ASIA motor score ismeasured on the two sides of the body in a group of 10key muscles, each of which is assigned a score between 0 and5 according to the Medical Research Council (MRC) musclegrades. The total ASIA motor score is the sum of these mea-surements and is an integer between 0 and 100. The sensoryexaminations assign to 27 index levels on each side betweenC3 and S4�5 a grade of 0, 1 or 2, according to whetherthere is absent, partial or normal response to light touchor pin prick. The total light touch or pin prick score is thusan integer between 0 and 108. One can also divide theseinto scores for the upper and lower extremities.

Table 6Marked recovery at Week 26 for cervical versus thoracic injuries withineach of three American Spinal Injury Association Impairment Scale sever-ity groups

Cervical Thoracic Significance

AIS A 49/317 (15.5%) 10/143 (7.0%) p�.02AIS B 48/103 (46.6%) 7/15 (46.7%) p�1.0AIS CD 106/126 (84.1%) 10/12 (83.3%) p�1.0

P values are from Fisher’s exact test.AIS�American Spinal Injury Association Impairment Scale.

All recruitment and measurement issues [1,2] were docu-mented extensively in our previous reports. Data were veri-fied onsite by a central team of monitors, and the databasewas checked. Statistical tests included Fisher’s exact testand by chi-squared (for nominal variables) and the nonpara-metric Wilcoxon two-sample test, normal approximation (forordinal variables).

Results

Recruitment was uneven, as shown in Table 4, and thedifferences among strata were highly statistically significant(p�.0001). Cervical trauma accounted for more than threequarters (579 of 760�76.2%) of the patients. Almost twothirds (482 of 760�63.4%) of the patients had complete (AISA) injuries. In contrast, there were few incomplete thoracicinjuries: 18 with AIS B and only 13 with AIS C or D.

Table 5 shows the performance of several potential prog-nostic factors as predictors of MR at Week 8 and at Week26. Two factors, injury region and severity, were dominantin attaining marked recovery. The cervical group did betterthan the thoracic (37.2% MR vs 15.9%, p�.0001). Also,AIS Groups C and D did better (p�.0001; 84.0% MR) thanGroup B (46.6%) which did better than Group A (12.8%).

In comparison, the prognostic value of other possiblestratification variables was generally modest. Having a sus-pected central cord injury, or one not involving a fracture

W.P. Coleman and F.H. Geisler / The Spine Journal 4 (2004) 373–378376

Table 7Motor and sensory recovery at Week 26 for cervical versus thoracic injuries within each of three American Spinal Injury Association ImpairmentScale severity groups

ASIA motor score difference C5–S1 Light touch score difference C3–S4�5 Pin prick score difference C3–S4�5

Cervical Thoracic p value Cervical Thoracic p value Cervical Thoracic p value

a. All measured spinal levelsAISA 9.6 4.7 .0001 13.8 5.3 .0004 8.3 4.0 .11B 31.7 21.0 .16 29.6 17.1 .20 27.5 17.5 .24CD 48.7 34.3 .002 24.7 19.1 .48 33.3 26.8 .46

b. Lower spinal levels

ASIA motor score difference L2–S1 Light touch score difference L1–S4�5 Pin prick score difference L1–S4�5

Cervical Thoracic p value Cervical Thoracic p value Cervical Thoracic p value

AISA 2.2 0.6 .01 4.5 1.6 .005 3.1 �2.7 .002B 19.3 18.9 .96 11.3 7.5 .61 15.5 13.1 .56CD 27.4 31.6 .30 9.0 9.2 .91 13.6 11.5 .94

P values are from the nonparametric Wilcoxon two-sample test, normal approximation.AIS�ASIA Impairment Scale; ASIA�American Spinal Injury Association.

dislocation, is significant (p�.001) at Weeks 8 and 26. How-ever, these two factors are forms of mild injury and areclosely related to AIS severity. The GM-1 drug effect issignificant (p�.01) at Week 8, but not Week 26. No othervariable analyzed (spinal surgery, surgical timing, MPSStiming, age or direct admission to tertiary care) was signifi-cant as a predictor of MR.

When both injury region and severity are considered to-gether in a two-way nominal logistic model, the whole modelis still highly significant (p�.0001), as is the term for injuryseverity factor; but neither the term for injury region (p�.37)nor its interaction with severity (p�.33) was nearly sig-nificant.

This suggests that part of the apparent recovery advantageof cervical over thoracic injuries may be the result of adifference in severity between these two groups. Table 4shows (p�.0001) that compared with the thoracic group,the cervical group had proportionally fewer patients withAIS A (332 of 579�57.3% vs 150 of 181�82.9%), whereasit had more with AIS B (113 of 579�19.5% vs 18 of181�9.9%) and still more with AIS CD (134 of 579�23.1%vs 13 of 181�7.1%).

Table 6 probes this further. Within AIS Group A the cervi-cal patients more often had more marked recovery (p�.02)

than did thoracic ones. However, the rates of MR were nearlyequal numerically and completely nonsignificant (p�1.0)between cervical and thoracic in the B and in the combined Cand D groups.

These observations are generally confirmed in Table 7a interms of ASIA motor score and light touch and pin prickscores. AIS Group A shows differences between cervicaland thoracic injuries in motor and light touch scores, butnot in pin prick. There are some numerical differences,but they are not significant (perhaps because of the smallsample size) in AIS Group B—or in Group CD, except formotor score.

In Table 7b, only the lower motor and sensory levelsare considered, in order to compare cervical and thoracicmore evenly. The results are generally as before, althoughshowing less trend among incomplete injuries toward in-creased recovery (and sometimes showing inferior) for thecervical group compared with the thoracic.

Thus, the better success in the cervical group seems tohave two main reasons: the fact that they had fewer completeinjuries and the fact that patients with complete cervicalinjuries did better than ones with complete thoracic injuries.

However, the cervical and thoracic groups showed verysimilar rates of marked recovery in patients with incomplete

Table 8The levels of the proposed injury region/severity variable

Number of patients at MR at Week 8 MR at Week 26Name Definition baseline in GM-1 study in GM-1 study in GM-1 study

a_c AIS grade A; cervical injury 332 7.8%, n�332 15.5%, n�317a_t AIS grade A; thoracic injury 150 2.0%, n�150 7.0%, n�143b AIS grade B; all injury regions 131 21.4%, n�131 46.6%, n�118cd AIS grade C and D; all injury regions 147 56.5%, n�147 84.1%, n�138

Differences between MR rates are significant (p�.0001).AIS�American Spinal Injury Association Impairment Scale; MR�marked recovery.

W.P. Coleman and F.H. Geisler / The Spine Journal 4 (2004) 373–378 377

injuries, although there was some trend for incomplete cervi-cal injuries to show larger recovery than incomplete thoracicinjuries in ASIA motor, light touch and pin prick scores.These later differences were not significant, very possiblybecause of the comparatively small sample size in thesegroups.

These observations, that there either is no difference be-tween the incomplete cervical and thoracic groups or else thatrecruitment is too small in those groups to positively sup-port such a conclusion, suggest a new stratification variable,“injury region/severity,” which is defined in Table 8. It distin-guishes cervical and thoracic injuries within AIS A but notin AIS B or AIS CD.

This new injury region/severity variable has reasonablesample size in each stratum while preserving detectable dif-ferences in marked recovery. It is a highly significant(p�.0001) predictor of MR, and the rates of MR at Week26 differ widely, ranging from 7.0% to 81.1%.

Table 9 shows the relationships of the injury region/severity variable to other predictors. It is most closely corre-lated with central cord injury, with fracture dislocation, withsurgical timing, and with age.

Discussion

These results are basic to our planned series of papersanalyzing the results of the GM-1 database, because theissues will arise in them continually. These same issues arelikely to arise in other acute spinal cord injury studies withsimilar target population and recruitment criteria. We hopethe results will be useful to others for planning and analysis.

The two variables, AIS severity and anatomical region ofinjury, both are strong (p�.0001) differentiators of patternsof patient recovery—more so than any other demographic,admission or treatment variables measured—and it is diffi-cult to assess intelligibly any of the other variables withouttaking differences in region and severity into account.

This fact is made more pressing by the imbalance inrecruitment. The current, very large trial, with 760 subjects,had only 18 thoracic injuries with AIS severity B, presentinga dilemma. Not to break one’s report down by region andseverity means extrapolating the results from large stratato the whole population, applying them equally to less-represented strata and thereby perhaps coming to clinicalconclusions that may not be justified. It would be equallymeaningless to use elaborate statistical models in an explicitattempt to differentiate treatment effects in such small strata.Although computer analysis software will deliver numbersreflecting purported results, they are found to be erratic andnot interpretable.

The injury region/severity variable maintains (p�

.0001) the prognostic value of using both region and severity,but it is simpler because it is a single factor with four levelsrather than a pair of factors: three severity grades crossedwith two anatomical regions. It is more statistically economi-cal. This variable gives a reasonable number of subjects ineach of its levels. As shown by the results in this paper, itis hard to avoid regarding its levels as individually homoge-neous, because either the incomplete cervical and thoracicgroups are the same or there is not enough sample size todistinguish them in practice.

Table 9Relation between injury region/severity variable and other therapeutic and prognostic indicators

Injury region/severity group

Category N Group a_c a_t b cd p Value

Central cord 69 Suspect 2.9% 0.0% 4.3% 92.8% �.0001691 Doubt or unknown 47.8% 21.7% 18.5% 12.0%

Type 72 Stable 25.0% 1.4% 15.3% 58.3% �.0001688 Fracture dislocation 45.6% 21.7% 17.4% 15.3%

Had spine surgery 600 Yes 45.2% 18.3% 18.0% 18.5% �.1160 No 38.1% 25.0% 14.4% 22.5%

Surgical timing group 25 �8 hours 52.0% 16.0% 16.0% 16.0% �.000594 8�x�24 49.0% 12.8% 19.2% 19.2%

174 24�x�72 50.6% 10.9% 24.1% 14.4%195 72�x�240 35.9% 29.2% 15.4% 19.5%112 240�x 48.2% 16.0% 12.5% 23.2%

Tertiary care admit 291 Direct 45.0% 20.3% 17.5% 17.2% .7469 Indirect 42.9% 19.4% 17.1% 20.7%

Age 367 �30 years 45.7% 23.2% 18.5% 13.1% �.0002393 �30 42.2% 16.5% 16.0% 25.2%

MPSS timing 469 �3 hours 45.8% 18.6% 18.3% 17.3% .1289 �3 hours 40.1% 21.8% 15.2% 22.8%

Treatment arm 430 GM-1 42.8% 19.3% 18.4% 19.5% .8330 Placebo 44.8% 20.3% 15.8% 19.1%

MPSS�methylprednisolone sodium succinate.

W.P. Coleman and F.H. Geisler / The Spine Journal 4 (2004) 373–378378

References

[1] Geisler F, Coleman W, Grieco G, et al. Recruitment and early treatmentin a multicenter study of acute spinal cord injury. Spine 2001;26(24S):58–67.

[2] Geisler F, Coleman W, Grieco G, et al. Measurement and recoverypatters in a multicenter study of acute spinal cord injury. Spine 2001;26(24S):68–86.

[3] Geisler F, Coleman W, Grieco G, et al. The Sygen multicenter studyof acute spinal cord injury. Spine 2001;26(24S):87–98.

[4] Hadley MN, Walters BC. Pharmacologic therapy following acute cervi-cal spinal cord injury. Neurosurgery 2002;50:S63–72.

Thirty YearsAgo inSpine

In 1974, Denis published the article thatpopularized the idea of judging post-traumatic stabilityof the thoraco-lumbar spine by considering the spine ascomposed of three columns: an anterior column of anteriorlongitudinal ligament, anterior half of the vertebral body,

[5] Geisler FH, Dorsey FC, Coleman WP. Recovery of motor function afterspinal-cord injury—a randomized, placebo-controlled trial with GM-1ganglioside. N Engl J Med 1991;324:1829–38.

[6] Schneider JS, Roeltgen DP, Mancall EL, Chapas-Crilly J, Rothblat DS,Tatarian GT. Parkinson’s disease: improved function with GM1 ganglio-side treatment in a randomized placebo-controlled study. Neurology1998;50(6):1630–6.

[7] American Spinal Injury Association, International Medical Societyof Paraplegia. International standards for neurologic and functionalclassification of spinal cord injury. Chicago, IL: ASIA/IMSOP,1992.

and anterior portion of the anulus fibrosus; a middlecolumn containing the posterior longitudinal ligament,posterior half of the vertebral body, and posterior half ofthe anulus; and a posterior column that includes theflavum, facet capsules, interspinous ligaments, and boneyelements of the neural arch [1]. The work modified theconcept of a two-column spine, popularized by Kelly andWhitesides [2].

References

[1] Denis F. Spinal instability as defined by the three-column spineconcept in acute spinal trauma. Clin Orthop 1984;189:65–76.

[2] Kelly RP, Whitesides TE Jr. Treatment of lumbodorsal fracture-dislocations. Ann Surg 1968;167:705–9.